基于改进Hopfield神经网络的对地攻击型无人机自主能力评价

对地攻击型无人机是当前最先进的无人装备之一，无人机必须具备很高的自主能力，自主能力成为无人机的典型作战能力。针对对地攻击型无人机的自主能力量化评价问题，从感知能力、决策能力、行为能力和安全能力4个方面，并侧重机载装备参数分析，提出了一套完整的自主能力评价指标体系。结合模型因素库，运用奇异值分解设计Hopfield神经网络权值矩阵，利用基于稀疏度的权值删减算法改进网络结构。构建自主能力评价标准，对对地攻击型无人机系统自主能力进行量化分级。仿真结果表明:相对于传统Hopfield神经网络，改进算法能够在一定范围内删除非关键的连接权值，降低网络复杂度，工程上更容易实现对地攻击型无人机系统自主能力的量化评价。 

Autonomous capability evaluation of ground-attack UAV based on improved Hopfield neural network

The ground-attack UAV has been one of the most state-of-the-art unmanned equipments, which requires a high degree of autonomous capability. Autonomous capabilityis a typical operational ability of UAV. In view of the quantitative evaluation of autonomous capability forground-attack UAV, this paper proposes a detailed evaluation index system of autonomous capability from four aspects of observation capability, decision capability, action capability and security capability, and places emphasis on the analysis of airborne equipment parameters. Combined with the model factor library, the weight matrix of Hopfield neural network is designed by singular value decomposition, and based on sparsity, the weight reduction algorithm is introduced to improve the network structure. Finally, the evaluation criterion of autonomy is established to quantify and grade the autonomous capability for ground-attack UAV system. The simulation results show that, compared with traditional Hopfield neural network, the improved algorithm can delete the unimportant connection weights within a certain range, reduce the network complexity, and easily achieve quantitative evaluation of the autonomous capability of UAV system.